Get started

Overview

What ActionWright is, what it can do on your PC, and what it needs.

What ActionWright is

ActionWright is a desktop assistant for Windows. You describe a task in plain words, typed or spoken, and ActionWright works on it on your PC. It can read and organize files in a folder you choose, look things up and use websites in its own browser, and operate desktop apps by looking at the screen.

It does the work in steps you can follow, asks you when something is missing, and waits for your approval before sensitive actions.

The parts of ActionWright

You talk to ActionWright through the task window or your voice. Behind them, a planner called the orchestrator splits your request into steps and hands each step to one of three agents.

PartWhat it does
Task windowThe main window, titled Agentic Task. You type a request, attach files, choose a folder, follow the progress, answer questions and read the result. Open it with ctrl+shift+a or from the tray icon. See Task window.
VoicePush-to-Talk is dictation: hold ctrl+space, speak, and ActionWright types what you said where your cursor is. Voice mode lets you say "Hey Yoo" to start a task and hear the answer. Voice mode is off by default. See Voice.
The orchestratorReads your request, plans the steps, looks up simple information on the web itself, asks you when it needs more detail, and hands each step to the right agent. It reports the result back to you. See How it works.
File agentFinds, reads, creates, edits, renames, moves and copies files. It works only inside the folder you choose for the task and its subfolders, or with files you attach. See Files.
Browser agentUses a separate browser made for ActionWright. Your normal browser and its sign-ins are not shared with it. When a site asks you to sign in, verify or solve a CAPTCHA, it pauses and asks you to take over. See Browser.
Computer UseLooks at what is on your screen and clicks, types and scrolls in desktop apps. It is on by default. You can turn it off in Settings with Enable Computer Use agent (screen control); the other agents keep working. See Computer Use.

What you need

You needDetails
WindowsWindows 10 version 2004 (build 19041) or later, or Windows 11. ActionWright is installed from the Microsoft Store.
Your own AI provider account and API keyActionWright uses an AI model from a provider you choose: OpenAI, xAI / Grok, Google / Gemini, Anthropic or OpenRouter, at the time of writing. You create the key in your provider account, and your provider bills you for the usage. See AI providers and keys.
An internet connectionAt launch, ActionWright checks that your version is still supported, and it does not start if it cannot. Every task also needs internet, because the AI model runs at your provider. See Required updates.
A microphone (optional)Only for Push-to-Talk and Voice mode. Everything else works without one.

What ActionWright is not

  • It is not a local AI model. The app runs on your PC, but the AI model runs at your provider. To work on a task, ActionWright sends your request and what the task needs to your provider: for example the text of documents you attach or it reads, the content of web pages, and screenshots of your screen or of its browser.
  • It does not include AI usage. It uses your key, so your provider bills you for every request a task makes. A long task makes many requests.
  • It is not a cloud service that works for you while your PC is off. Tasks run while ActionWright is running on your PC.
  • It does not send your tasks to ActionWright's own service. Only the launch version check contacts that service, and your requests, files and screenshots are not sent there. See Required updates.
  • It is not free of mistakes. AI models can misread a request or a page. Read approval dialogs before you answer them, and check results that matter. See Safety.

How a task flows

  1. You type a request in the task window, or say it after "Hey Yoo" if Voice mode is on. For file work you attach files or choose a root folder.
  2. The orchestrator plans the steps. For a simple question it may search the web and read pages itself.
  3. It hands steps to the file agent, the browser agent or Computer Use. Each step appears in the Execution Trace panel of the task window.
  4. If it needs more detail, it asks you in the task window. Before a sensitive action, it shows an Action Approval Required dialog where you choose Allow Once or Deny. Some actions also offer Always Allow.
  5. If a task uses up its step budget, it pauses and asks whether to continue. You choose Yes, Continue or No, Stop.
  6. The answer appears under Result. If the task window is not in front, Windows shows a Task complete notification.

You can stop a task at any time with the Stop button or ctrl+esc. The action that is running finishes first, then the task stops. For the full picture, read How it works.

Where to go next